import pandas as pd import math from functools import reduce from odps import ODPS from threading import Timer from datetime import datetime, timedelta from get_data import get_data_from_odps from db_helper import RedisHelper from utils import filter_video_status, check_table_partition_exits from config import set_config from log import Log config_, _ = set_config() log_ = Log() features = [ 'apptype', 'videoid', 'preview人数', # 过去24h预曝光人数 'view人数', # 过去24h曝光人数 'play人数', # 过去24h播放人数 'share人数', # 过去24h分享人数 '回流人数', # 过去24h分享，过去24h回流人数 'preview次数', # 过去24h预曝光次数 'view次数', # 过去24h曝光次数 'play次数', # 过去24h播放次数 'share次数', # 过去24h分享次数 'platform_return', 'platform_preview', 'platform_preview_total', 'platform_show', 'platform_show_total', 'platform_view', 'platform_view_total', ] def data_check(project, table, now_date): """检查数据是否准备好""" odps = ODPS( access_id=config_.ODPS_CONFIG['ACCESSID'], secret_access_key=config_.ODPS_CONFIG['ACCESSKEY'], project=project, endpoint=config_.ODPS_CONFIG['ENDPOINT'], connect_timeout=3000, read_timeout=500000, pool_maxsize=1000, pool_connections=1000 ) try: dt = datetime.strftime(now_date, '%Y%m%d') check_res = check_table_partition_exits(date=dt, project=project, table=table) if check_res: sql = f'select * from {project}.{table} where dt = {dt}' with odps.execute_sql(sql=sql).open_reader() as reader: data_count = reader.count else: data_count = 0 except Exception as e: data_count = 0 return data_count def get_feature_data(now_date, project, table): """获取特征数据""" dt = datetime.strftime(now_date, '%Y%m%d') records = get_data_from_odps(date=dt, project=project, table=table) feature_data = [] for record in records: item = {} for feature_name in features: item[feature_name] = record[feature_name] feature_data.append(item) feature_df = pd.DataFrame(feature_data) return feature_df def cal_score(df, param): # score计算公式: score = share次数/(view+1000)+0.01*return/(share次数+100) df = df.fillna(0) if param.get('view_type', None) == 'video-show': df['share_rate'] = df['share次数'] / (df['platform_show'] + 1000) elif param.get('view_type', None) == 'preview': df['share_rate'] = df['share次数'] / (df['preview人数'] + 1000) else: df['share_rate'] = df['share次数'] / (df['view人数'] + 1000) df['back_rate'] = df['回流人数'] / (df['share次数'] + 100) df['score'] = df['share_rate'] + 0.01 * df['back_rate'] df['platform_return_rate'] = df['platform_return'] / df['回流人数'] df = df.sort_values(by=['score'], ascending=False) return df def video_rank_h(df, now_date, rule_key, param, data_key): """ 获取符合进入召回源条件的视频，与每日更新的rov模型结果视频列表进行合并 :param df: :param now_date: :param rule_key: 天级规则数据进入条件 :param param: 天级规则数据进入条件参数 :param data_key: 使用数据标识 :return: """ redis_helper = RedisHelper() log_.info(f"videos_count = {len(df)}") # videoid重复时，保留分值高 df = df.sort_values(by=['score'], ascending=False) df = df.drop_duplicates(subset=['videoid'], keep='first') df['videoid'] = df['videoid'].astype(int) day_recall_videos = df['videoid'].to_list() log_.info(f'day_by30day_recall videos count = {len(day_recall_videos)}') # 视频状态过滤 filtered_videos = filter_video_status(day_recall_videos) log_.info('filtered_videos count = {}'.format(len(filtered_videos))) # 获取top视频 top = param.get('top') day_recall_df = df[df['videoid'].isin(filtered_videos)] day_recall_df = day_recall_df.sort_values(by=['score'], ascending=False) day_recall_df = day_recall_df[:top] # 写入对应的redis now_dt = datetime.strftime(now_date, '%Y%m%d') day_video_ids = [] day_recall_result = {} for video_id in day_recall_df['videoid'].to_list(): score = day_recall_df[day_recall_df['videoid'] == video_id]['score'] day_recall_result[int(video_id)] = float(score) day_video_ids.append(int(video_id)) day_30day_recall_key_name = f"{config_.RECALL_KEY_NAME_PREFIX_30DAY}{data_key}:{rule_key}:{now_dt}" if len(day_recall_result) > 0: log_.info(f"count = {len(day_recall_result)}, key = {day_30day_recall_key_name}") redis_helper.add_data_with_zset(key_name=day_30day_recall_key_name, data=day_recall_result, expire_time=2 * 24 * 3600) def merge_df(df_left, df_right): """ df按照videoid 合并，对应特征求和 :param df_left: :param df_right: :return: """ df_merged = pd.merge(df_left, df_right, on=['videoid'], how='outer', suffixes=['_x', '_y']) df_merged.fillna(0, inplace=True) feature_list = ['videoid'] for feature in features: if feature in ['apptype', 'videoid']: continue df_merged[feature] = df_merged[f'{feature}_x'] + df_merged[f'{feature}_y'] feature_list.append(feature) return df_merged[feature_list] def merge_df_with_score(df_left, df_right): """ df 按照videoid合并，平台回流人数、回流人数、分数 分别求和 :param df_left: :param df_right: :return: """ df_merged = pd.merge(df_left, df_right, on=['videoid'], how='outer', suffixes=['_x', '_y']) df_merged.fillna(0, inplace=True) feature_list = ['videoid', '回流人数', 'platform_return', 'score'] for feature in feature_list[1:]: df_merged[feature] = df_merged[f'{feature}_x'] + df_merged[f'{feature}_y'] return df_merged[feature_list] def rank(now_date, rule_params, project, table): # 获取特征数据 feature_df = get_feature_data(now_date=now_date, project=project, table=table) feature_df['apptype'] = feature_df['apptype'].astype(int) # rank data_params_item = rule_params.get('data_params') rule_params_item = rule_params.get('rule_params') for param in rule_params.get('params_list'): score_df_list = [] data_key = param.get('data') data_param = data_params_item.get(data_key) log_.info(f"data_key = {data_key}, data_param = {data_param}") rule_key = param.get('rule') rule_param = rule_params_item.get(rule_key) log_.info(f"rule_key = {rule_key}, rule_param = {rule_param}") merge_func = rule_param.get('merge_func', 1) if merge_func == 2: for apptype, weight in data_param.items(): df = feature_df[feature_df['apptype'] == apptype] # 计算score score_df = cal_score(df=df, param=rule_param) score_df['score'] = score_df['score'] * weight score_df_list.append(score_df) # 分数合并 df_merged = reduce(merge_df_with_score, score_df_list) # 更新平台回流比 df_merged['platform_return_rate'] = df_merged['platform_return'] / df_merged['回流人数'] video_rank_h(df=df_merged, now_date=now_date, rule_key=rule_key, param=rule_param, data_key=data_key) else: df_list = [feature_df[feature_df['apptype'] == apptype] for apptype, _ in data_param.items()] df_merged = reduce(merge_df, df_list) score_df = cal_score(df=df_merged, param=rule_param) video_rank_h(df=score_df, now_date=now_date, rule_key=rule_key, param=rule_param, data_key=data_key) def rank_bottom(now_date, rule_params): """未按时更新数据，用前一天数据作为当前的数据""" redis_helper = RedisHelper() redis_dt = datetime.strftime(now_date - timedelta(days=1), '%Y%m%d') key_prefix_list = [config_.RECALL_KEY_NAME_PREFIX_30DAY] for param in rule_params.get('params_list'): data_key = param.get('data') rule_key = param.get('rule') log_.info(f"data_key = {data_key}, rule_key = {rule_key}") for key_prefix in key_prefix_list: key_name = f"{key_prefix}{data_key}:{rule_key}:{redis_dt}" initial_data = redis_helper.get_all_data_from_zset(key_name=key_name, with_scores=True) if initial_data is None: initial_data = [] final_data = dict() for video_id, score in initial_data: final_data[video_id] = score # 存入对应的redis final_key_name = \ f"{key_prefix}{data_key}:{rule_key}:{datetime.strftime(now_date, '%Y%m%d')}" if len(final_data) > 0: redis_helper.add_data_with_zset(key_name=final_key_name, data=final_data, expire_time=2 * 24 * 3600) def timer_check(): project = config_.PROJECT_30DAY_APP_TYPE table = config_.TABLE_30DAY_APP_TYPE rule_params = config_.RULE_PARAMS_30DAY_APP_TYPE now_date = datetime.today() log_.info(f"now_date: {datetime.strftime(now_date, '%Y%m%d')}") now_h = datetime.now().hour # 查看当前天级更新的数据是否已准备好 data_count = data_check(project=project, table=table, now_date=now_date) if data_count > 0: log_.info(f'day_by30day_data_count = {data_count}') # 数据准备好，进行更新 rank(now_date=now_date, rule_params=rule_params, project=project, table=table) elif now_h > 3: log_.info('day_by30day_recall data is None!') rank_bottom(now_date=now_date, rule_params=rule_params) else: # 数据没准备好，5分钟后重新检查 Timer(5 * 60, timer_check).start() if __name__ == '__main__': timer_check()